Claude spent 2 days reimplementing vLLM — readable code review reveals a maintainability nightmare
bingxu_ · x · 2026-10-23
bingxu showcases int21ai's AI-generated Rust inference factory (claims 1700+ tokens/s on DeepSeek v4.1 Flash, beating vLLM/SGLang on 20+ models), with a screenshot of Gregory Dianos reviewing Claude's 2-day reimplementation of vLLM, Megatron-core and ScalarLM: stray ENV variables, a casual malloc reimplementation, 15 unorganized prototype kernels, an invented decorator bypassing MLIR types, three scattered matmul implementations. His verdict: promising but an unmaintainable nightmare if unread — reports of software engineering's demise are premature.
More from coding & agent
- GitHub commits jumped from 1B to 14B in a year, but LOC is not productivity — srchvrs · 2026-10-23
- Agent Buddy, the ESP32 desk robot for coding agents, gets a community fork — DanWahlin · 2026-10-23
- Grok Build ships v1.0.66/67: @ file completion 8x faster, worktree approvals — mark_k · 2026-10-23
- Prime Intellect lays out continual learning roadmap: mining agent exhaust for self-improving models — willcb · 2026-10-23
- Percy Liang teaches new MIT course on self-evolving AI agents across four evolution levels — pliang279 · 2026-10-23
- Using GrokBot as an OSINT Tool to Trace the Origin of Forged AfD Emails — FlorianGallwitz · 2026-10-23